edit: ROFL thanks for moving this, no clue how i missed this thread

found this thread digging through the archives. Has anyone in here played around w/ neural networks and NT thru matlab or by using FANN?
Neural Networks - [AUTOLINK]NinjaTrader[/AUTOLINK] Support Forum

tortexal 

found this thread digging through the archives. Has anyone in here played around w/ neural networks and NT thru matlab or by using FANN? Neural Networks - [AUTOLINK]NinjaTrader[/AUTOLINK] Support Forum

I use matlab with TWS (interactive brokers java client) occasionally, it is great for prototyping ideas if you are familiar with matlab already, otherwise I wouldn't waste time with it.

here is another site which has some Neural net resources
Resources for Neural Networks and Knowledge Modeling

and here is a link to a site about gaussian process regression which is another machine learning technique (the site also has info which is useful for neural networks and other probabilistic modeling techniques)
The Gaussian Processes Web Site

I would highly advise anyone interested in implementing NN or any other machine learning technique to read the preface/intro sections of the free ebook here:
Gaussian Processes for Machine Learning: Contents

They have a very insightful overview of the machine learning landscape including the pros and cons of different techniques. Particularly of interest to this topic is that they discuss how most researchers in machine learning have given up on the idea of supervised neural networks (ie those trained by backpropagation) for solving complex problems.

I do not know whether it is helpful, here's a book for matlab and c #

https://www.lephanpublishing.com/MATLABbook/MatlabCsharpSecond/MATLABBookCsharpSecond.html

sefstrat,

do you use the levenberg marquardt alg. in your nn?

wh

wh 

sefstrat, do you use the levenberg marquardt alg. in your nn? wh

Yes, I use the implementation from alglib

Anybody using R?

yes but not for trading only for geostatistic with rserve and java 4 web application ... i think the integration with c# and r is to complicated (for me). i think matlab is much easier ... and i found today a new nice tool but not tested <google>findgraph</google> ... with some good features

wh

wh in your opinion is it worth the time for a guy/gal retail investor who likes math and strat trading to learn R/S/Matlab over and above Ninja/C#? If you were starting from scratch where would you spend your time?

Thanks for your input. It is appreciated. I'm blown away by some of the stuff posted here.

MXASJ 

wh in your opinion is it worth the time for a guy/gal retail investor who likes math and strat trading to learn R/S/Matlab over and above Ninja/C#? If you were starting from scratch where would you spend your time? Thanks for your input. It is appreciated. I'm blown away by some of the stuff posted here.

that is not so easy to answer. if you can an object-oriented language like c + +, Java or C #, so will look into matlab. the educational environment looks better then r or s plus. so you can start with statistic. if you are not a statistic so i would look into c # and ninja script api to understand the main concept of both. later you will come back to statistic.

I come from the geostatistic and had only with points and polygons to do and now it is time series analysis. so i must learn as you.

better hold or keep it simple, and it will come.

wh

wh 

that is not so easy to answer. if you can an object-oriented language like c + +, Java or C #, so will look into matlab. the educational environment looks better then r or s plus. so you can start with statistic. if you are not a statistic so i would look into c # and ninja script api to understand the main concept of both. later you will come back to statistic. I come from the geostatistic and had only with points and polygons to do and now it is time series analysis. so i must learn as you. better hold or keep it simple, and it will come. wh

Well, the jury is in !!!!!

I now know that I will never be more than a HotShot tweaker.

If someone else creates a program, I can turbocharge its performance.

Man's got to know his limitations...

RJay